Revolut has launched Revolut Research, a dedicated division within its broader AI Department focused on developing next-generation machine learning architecture for financial services.
With more than 80 million retail customers across over 40 markets, Revolut said the new division will provide a unified foundation for its proprietary AI deployments and machine learning initiatives. The unit will work alongside leading academic and technology institutions to advance financial AI capabilities.
Revolut Research serves as the engine behind PRAGMA, the company’s proprietary foundation model developed in collaboration with NVIDIA. The model is designed specifically for financial applications, including real-time risk assessment, fraud detection, platform operations and personalised product recommendations.
Early deployments of PRAGMA using historical data have reported significant improvements compared with legacy baselines. Revolut said the model achieved 2.3 times higher accuracy in identifying credit default risk, detected 65% more fraud cases while improving alert precision by 17%, and generated 41% more relevant product recommendations across retail and business accounts.
The company attributes part of this performance to its extensive global dataset. Revolut processes billions of transactions across more than 40 markets, generating diverse financial behaviour data that can be used to train and improve its AI models.
Revolut said this creates a continuous feedback loop in which expanding data can strengthen the models’ ability to identify fraudulent activity, assess financial risk and anticipate customer needs.
The company is positioning the approach as part of its broader “build, don’t bolt on” strategy, with AI capabilities being developed directly within its financial technology infrastructure rather than relying solely on third-party software.
Beyond PRAGMA, Revolut’s security models review nearly one billion transactions every month to proactively identify fraudulent activity. Its AI capabilities also support customer-facing tools such as AIR, an in-app assistant capable of executing complex financial tasks in a single step, currently available in the UK.
Revolut Research is led by Anton Repushko, who has presented the division’s research at international technology and machine learning conferences, including ICML in Seoul.
Looking ahead, Revolut Research plans to publish scientific findings and open-source technical frameworks while engaging with the wider research community through international conferences and quarterly scientific meet-ups.
The launch signals Revolut’s growing focus on proprietary AI infrastructure and financial-specific machine learning, with the company seeking to use its global operational data and technology ecosystem to develop more intelligent, secure and personalised banking services.
